Rescale’s Post

Transcript

Alice's work is done on the optimization of the heat exchanger. But something in her head tells her that it can be better. Rather than continuing to iterate with expensive physics-based simulations, she decides to train an AI model using the data she generated in the study. As you can see, she has a bunch of design directly in the study. So what she does is she trains this AI model by simply going into the AI tab within the Rescale study. This turns every simulation engineer into a data scientist with the ability to leverage the power of AI. Alice goes to the study, clicks on the AI tab, which guides her through creating an AI model using her engineering knowledge rather than data science knowledge. She leverages the heat exchanger simulation data to train the model with the goal to further optimize our model. This process first processes the simulation data, second, trains the model, thirdly, deploys the model, and then fourth, as new simulation data becomes available, the model is automatically fine tuned. This is then added back into the software catalog as a tile for Alice to leverage. So once the model is trained, Alice can go in and use this model to optimize her heat exchanger, either going forwards or backwards by using the input or output parameters as the goal. To do so, she clicks on new job. As she creates a new job, you'll see we can create an AI simulation. I'll go into the software where we'll add AI, and then what we'll see is we'll be able to leverage the newly created model that we just built. In this case, we're using Rescale's in- house solver, Rescale AI Physics software. She can use these in jobs or workstations or visualize them directly. And once AI predictions are generated, validation is the key to ensuring accuracy. So Alice runs the simulation, finds an optimization, then reruns her COMSOL simulations to validate that the simulation is accurate. And also, AI models are also used for not only design optimization, but also for maintenance in the field. So as soon as we release this heat exchanger to the field, we'll keep the AI model for any maintenance or validation going forward.